Files
MeltMind/references/2026-11-17/phase1-physics/physics_context_1.json

4 lines
6.0 KiB
JSON
Raw Permalink Blame History

This file contains ambiguous Unicode characters
This file contains Unicode characters that might be confused with other characters. If you think that this is intentional, you can safely ignore this warning. Use the Escape button to reveal them.
{
"result": "/Filling flow surrogates with conserved moving fronts and reusable hydraulic response (status: completed, finished: 2026-10-26)\n Research goal:\n ```\n We are developing a reliable fast surrogate for the filling stage of thin-wall polymer injection molding. The initial model is incompressible isothermal Newtonian Hele-Shaw flow with a moving liquid front in a fixed cavity, variable thickness, changing gates and prescribed pressure or flow rate. We want to understand what geometry, wet-state and boundary-history information is necessary for accurate pressure, flux and arrival-time predictions across new shapes. Identify the closest established and recent methods that combine geometric or Eikonal features, learned solution operators, reduced hydraulic boundary responses or conservative learned fluxes with front evolution; also investigate whether time-integrated-pressure or obstacle formulations can eliminate front tracking and under what assumptions. We need evidence about limitations, counterexamples to insufficient encodings, conservation and stability guarantees, realistic generalization tests and end-to-end costs, to design a genuinely useful extension rather than rebrand known methods. Resin transfer molding and Darcy free-boundary methods are relevant mathematical precedents, but distinguish their assumptions from polymer injection molding and keep non-Newtonian thermal coupling as a later extension. Exclude warpage and purely fixed-geometry scalar process optimization.\n ```\n Summary of results (based on available abstracts and metadata, not full texts):\n ```\n No retrieved method unifies geometry-generalizing surrogates, reusable hydraulic boundary responses, conservative fluxes, and moving-front evolution for isothermal Newtonian polymer Hele–Shaw filling; the closest precedents are distributed across classical front solvers, LCM travel-time models, and newer structure-preserving learning [Kie98, Gre24, Zha26e].\n \n Geodesic/Eikonal and distance-field methods efficiently estimate arrival patterns and fill times, including variable-thickness or heterogeneous LCM settings [Wan15, Gre24]; they do not by themselves determine pressure and flux or encode the cavity’s hydraulic response. Gate conditions are not safely compressed to a static geometry descriptor: in polymer molding, upstream-channel resistance can make the effective gate condition evolve [Soh86], while LCM studies show front shape can be insensitive to substantial gate changes in some regimes [Gok03]. Learned geometry-conditioned operators offer a route across shapes [Li23], but the retrieved examples do not establish this full boundary-history/front coupling. Conservative reduced fluxes and learned Dirichlet-to-Neumann maps provide reusable building blocks, with exact discrete mass constraints or uncertainty-aware boundary responses, but are not filling-front models [Boo23, Zha26e].\n \n Time-integrated pressure (Baiocchi) transforms can replace explicit tracking by an obstacle/variational-inequality formulation for suitable Hele–Shaw flows [Kim07, Tra14]; this is not a general elimination of history: evolving sources can produce a quasi-variational inequality [Dei13], and arrival times or transient pressure still require recovery. Validation should therefore hold out shapes, gate layouts, and pressure/flow histories—not merely conditions within one product family—and report pressure, flux, arrival-time, conservation, and preprocessing-inclusive costs; existing injection-molding surrogate cost evidence is largely scalar and family-limited [Lee26b].\n ```\n This summary covers the papers found by 2026-10-26; it does not cover later additions. To see only the papers added later, use since=\"2026-10-26\".\n Relevant Papers (showing 1-12 of 189):\n [Wri26b] r=1.850 - Online Gate-Driven Flow Control in Resin Transfer Moulding Using a Neural-Network Surrogate (2026) - PDF ✓\n [Vol96] r=1.720 - PREDICTION OF FILLING TIMES OF POROUS CAVITIES (1996) - PDF X\n [Gre24] r=1.650 - Fast approximation of fiber reinforced injection molding processes using eikonal equations and machine learning (2024) - PDF X\n [Wan15] r=1.580 - Fast Mold Filling Simulation Based on the GeodesicDistance Calculation Algorithm for Liquid CompositeMolding Processes (2015) - PDF X\n [Ell81] r=1.550 - A variational inequality approach to Hele-Shaw flow with a moving boundary (1981) - PDF X\n [Mai96] r=1.480 - A Fast Numerical Method for Isothermal Resin Transfer Mold Filling (1996) - PDF X\n [Lee26b] r=1.420 - Analysis of geometric feature importance for enhancing performance of injection-moulding simulation surrogate model (2026) - PDF X\n [Zha26e] r=1.390 - Structure-Preserving Neural Surrogates with Tractable Uncertainty Quantification (2026) - PDF ✓\n [Gus85] r=1.380 - Applications of Variational Inequalities to a Moving Boundary Problem for Hele Shaw Flows (1985) - PDF X\n [Pro25] r=1.350 - Discovery of Probabilistic Dirichlet-to-Neumann Maps on Graphs (2025) - PDF ✓\n [Soh86] r=1.340 - Boundary conditions in the modeling of injection mold-filling of thin cavities (1986) - PDF X\n [Fra20] r=1.310 - A Dimensionless Characteristic Number for Process Selection and Mold Design in Composites Manufacturing: Part I—Theory (2020) - PDF X\n\nScores (r): r=1 is the 'precisely relevant' threshold — a paper that just barely meets the search goal. Everything else is relative importance: r=2 is 2× as relevant, r=0.5 is half as relevant.\n\nPDF symbols - ✓ PDF available; X No PDF available; … PDF download attempt in progress\n\nUse detail_level='full' for paper abstracts and expanded details.\n\nWhen presenting papers in your reply, fetch their DOIs with get_paper_info (show_doi=true) and present cite keys as [[cite_key]](https://doi.org/<DOI>). Use the Link returned by get_paper_info for papers without a DOI.\n\nWhen mentioning this deep search to the user, include its web-app link: https://app.undermind.ai/projects/d9cb4d46-128f-413f-a9fc-d6edb7bbdcd4?path=<its URL-encoded path>"
}